why we filter smps

There's a guy giving away free PCBs for a 3 Ampere (or less), 48 Volt (or less), SMPS filter. It's aimed at analog line level equipment, and the title of the thread says it's intended application is "Preamps, HeadPhone Amps, Korg NuTube, etc."

His generosity is not completely unlimited, of course; have a look at post #399 (here)
 
For a DAC, it could cause intermodulation between images of the signal and SMPS harmonics, for example.

If it crosstalks to the DAC's reference voltage, the DAC will multiply the ripple with the discrete-time audio signal. The discrete-time audio signal has copies of the audio spectrum (images) around all multiples of the sample rate. The multiplying with the ripple will convert some of these images into the audio band in much the same way as the RF signal is converted to IF by the mixer stage in a radio.

This doesn't matter if the ripple is at an exact multiple of the sample rate, because the converted image is then just a copy of the desired audio signal, but if the ripple is not at an exact multiple of the sample rate, it will cause a frequency-shifted version of the audio to be added to the desired signal.

If so why any linear / regulated power supply design do not use common mode filters, ferrite beads and so on.

Current linear power supply design focus on lowering the ripple and then use state of art regulators which is so weak / none attenuation performance at high frequencies.

For the linear power supplies there are lots of high frequency content comes from mains and injected in to secondary of the transformer.

If we look at the very top performer dac designs, they do not use any mains emi filter. But still have great objective measurements.
 
Compared to a switching MOSFET drain jumping up and down between 0 and 300 V at a 100 kHz or so rate, I think the amount of high-frequency rubbish coming from the mains is not that high.

In any case, in the mixed-signal circuits I worked on, there were usually ferrite beads on the boards to separate the digital from the analogue and the supplies from the analogue. Usually these are SMD ferrite beads like the Murata BLM18 series placed right next to sensitive and noisy supply pins.
 
Marcel: How would you filter SMPS noise that sounds like a blend of 120Hz and "clock"? 12V 20A max output. By clock I mean it's not a constant HF hash, but you here the feedback "clocking" like when you hear your computer processing through a cheap sound card.
 
It supplies tube heaters, DC-DC boost for B+, and I only hear noise on a high gain low PSRR stage (line stage is quiet even on headphones, but ad 40db and voila). It sounds like 120Hz hum plus HF clock style hash? I haven't tried to filter it at all save for normal RC filtering at the gain stages.
 
Last edited:
I guess I would first want to figure out whether there is audible-frequency-range interference present at the SMPS output, or if it is some interaction with the boost converter or a high-frequency ripple with varying amplitude that becomes audible due to parasitic AM demodulation.

You could AC couple the supplies into the input of an amplifier, using diodes to prevent blowing up the amplifier at power-on or -off and using an RC low-pass at 20 kHz. If you hear your hum blend at the 12 V under those conditions, it is apparently due to poor regulation of the switching regulator. A series regulator between the SMPS and anything sensitive should then help. If not, then a filter like dotneck335 referred to might help, although you have to redimension it if it has to handle 20 A.

The fact that you only hear it on a low PSRR stage suggests it goes via the boost converter and B+. If that's correct, you could leave the heater supply untouched.
 
Last edited:
If so why any linear / regulated power supply design do not use common mode filters, ferrite beads and so on.

Current linear power supply design focus on lowering the ripple and then use state of art regulators which is so weak / none attenuation performance at high frequencies.

For the linear power supplies there are lots of high frequency content comes from mains and injected in to secondary of the transformer.

If we look at the very top performer dac designs, they do not use any mains emi filter. But still have great objective measurements.

Then I must have seen a Fata Morgana 🙂 Many new designs use RF protection and many recent regulators are meant for use in RF environments (i.e. cell phone, network devices).

Personally I use beads and common mode filters in linear designs for a very long time. Many quality brands do the same although it may not look like that at first (built in EMI filters). The use of EMI filters at the mains side is meanwhile a necessity when one lives in a city with industry, uses a cell phone, has wireless internet, bluetooth, LED lighting, SMPS adapters for many devices in a home etc.

Choice of mains transformer makes a difference as well.
 
Last edited:
According to very detailed test reports here below;
PS Audio Noise Harvester AC Cleaner Review | Audio Science Review (ASR) Forum
and
Review and Measurements of Schiit Yggdrasil V2 DAC | Page 15 | Audio Science Review (ASR) Forum

They simply shows that emi/diff/cm noises which are mainly above 10Mhz do not have any negative effect on such sensitive device like DAC.

I don't think we agree what a mains filter is. I took the time to read the first link and that is no mains filter, that is an plug in audio gadget for people with too much money. It does not seem to do something really useful except blinking a LED which is absolutely not exemplary for real EMI mains filters (these have no LED). So this example proves nothing. Please read something about EMI mains filters and what they do. These are used in industrial (and audio) applications a lot with a reason. The second link has a lot of debating/discourse but apart from 2 measured DACs of which one clearly has PSU issues I can't conclude your remark " It simply shows that emi/diff/cm noises which are mainly above 10Mhz do not have any negative effect on such sensitive device like DAC."

Measurements are nice but beware to have opinions based on just a few tests of a few devices certainly when audio gadgets and a flawed device are used. In both links there is no measurement of an EMI Mains filter.
 
Last edited:
I put the choke on the B+ of the boosted 280V supply, not the 12V. It was a wash anyway, what works on the bench doesn't work on the rack. Same clock noise nonsense. Now I'll swap in a passive supply to see if it's RF or something. I can't receive AM radio in this place.